Massively Parallel Methods for Deep Reinforcement Learning

About

We present the first massively distributed architecture for deep reinforcement learning. This architecture uses four main components: parallel actors that generate new behaviour; parallel learners that are trained from stored experience; a distributed neural network to represent the value function or behaviour policy; and a distributed store of experience. We used our architecture to implement the Deep Q-Network algorithm (DQN). Our distributed algorithm was applied to 49 games from Atari 2600 games from the Arcade Learning Environment, using identical hyperparameters. Our performance surpassed non-distributed DQN in 41 of the 49 games and also reduced the wall-time required to achieve these results by an order of magnitude on most games.

Arun Nair, Praveen Srinivasan, Sam Blackwell, Cagdas Alcicek, Rory Fearon, Alessandro De Maria, Vedavyas Panneershelvam, Mustafa Suleyman, Charles Beattie, Stig Petersen, Shane Legg, Volodymyr Mnih, Koray Kavukcuoglu, David Silver · 2015

Related benchmarks

Task	Dataset	Result
Reinforcement Learning	Atari 2600 MONTEZUMA'S REVENGE	Score84	45
Reinforcement Learning	Atari 2600 57 games (test)	Median Human-Normalized Score—	15
Atari Game Playing	Atari 2600 57 games human starts evaluation metric	Median Human-Normalized Score71.3	14
Reinforcement Learning	Atari 2600 Arcade Learning Environment (evaluation)	Montezuma's Revenge Score4	11
Game Playing	Atari 2600 human starts 49 games (test)	Median Normalized Score47.5	3
Atari games	Atari 2600 49 games, no-op starts (test)	Median Normalized Performance93.5	2